Stubbs, F. E., Birnie, M. T., Biddie, S. C., Lightman, S. L., & Conway- Campbell, B. L. (2018). SKOV3 cells containing a truncated ARID1a have a restricted genome-wide response to glucocorticoids. Molecular and Cellular Endocrinology, 461, 226-235. https://doi.org/10.1016/j.mce.2017.09.018

Peer reviewed version License (if available): CC BY-NC-ND Link to published version (if available): 10.1016/j.mce.2017.09.018

Link to publication record in Explore Bristol Research PDF-document

This is the accepted author manuscript (AAM). The final published version (version of record) is available online via Elsevier at https://doi.org/10.1016/j.mce.2017.09.018 . Please refer to any applicable terms of use of the publisher.

University of Bristol - Explore Bristol Research General rights

This document is made available in accordance with publisher policies. Please cite only the published version using the reference above. Full terms of use are available: http://www.bristol.ac.uk/red/research-policy/pure/user-guides/ebr-terms/ Elsevier Editorial System(tm) for Molecular and Cellular Endocrinology Manuscript Draft

Manuscript Number: MCE-D-17-00106R1

Title: SKOV3 cells containing a truncated ARID1a protein have a restricted genome-wide response to glucocorticoids

Article Type: Research Paper

Keywords: Glucocorticoids, Glucocorticoid , SWI/SNF, ARID1a, .

Corresponding Author: Dr. Becky L Conway-Campbell,

Corresponding Author's Institution: University of Bristol

First Author: Felicity E Stubbs, PhD

Order of Authors: Felicity E Stubbs, PhD; Matthew T Birnie, B.Sc. ; Simon C Biddie, PhD; Stafford L Lightman, PhD; Becky L Conway-Campbell

Abstract: AT-rich interacting domain subunit 1a (ARID1a) is an essential SWI/SNF component frequently mutated in human cancers. ARID1a mutations have also been associated with glucocorticoid resistance, potentially related to the well-established role of the SWI/SNF complex in glucocorticoid target regulation. Glucocorticoids are steroid hormones important for regulating many physiological processes through the activation of the (GR). As GR interacts directly with ARID1a, we hypothesized that a truncating ARID mutation would interfere with GR-dependent gene regulation. Using high throughput RNA sequencing (RNA-SEQ) we show a restricted glucocorticoid response in SKOV3 cells, which contain an inactivating ARID1a mutation. We also show a lack of GR binding at the GR-dependent regulatory site in the Period 1 gene, which has previously been shown to require remodelling. Taken together, our data suggests that ARID1a may be required for regulation of a subset of glucocorticoid responsive . In the case of SKOV3 cells, in which ARID1a is mutated, glucocorticoid-dependent transcriptional regulation of these genes is significantly impaired.

Cover Letter

Dr BL Conway-Campbell Research Fellow Dorothy Hodgkin Building Whitson St. Bristol, BS1 3NY Tel: +44 (0)117 3313138 Fax: +44 (0)117 3313035 [email protected]

06 August 2017

Dear Editor

We thank you and your reviewers for your extremely insightful and kind comments about our manuscript. We have attempted to address all the recommendations in our revised manuscript. Unfortunately, all our attempts to transfect the SKOV cells (for either reintroduction of full length wild type ARID1a or knock-down of the existing ARID1a) proved impossible. Therefore, we have shown all the data obtained from our attempts to do as the reviewer asked, in our response to reviewers letter (Appendices). We are confident that with our revisions in text to clarify certain points, and include suggested discussion, that our manuscript would be of interest to your journal’s audience. Therefore, we would be very pleased if you would consider our revised manuscript entitled ‘SKOV3 cells containing a truncated ARID1a protein have a restricted genome-wide response to Glucocorticoids’ for publication in Molecular and Cellular Endocrinology. We believe this work would still be of interest to those focused on ARID1a mutations in cancer as well as to glucocorticoid researchers.

Yours faithfully,

Becky Conway-Campbell

*Revision Notes

1. Formally, genetic manipulation of ARID1a (either rescue by wt protein overexpression or knockdown of the remaining wt protein) in SKOV3 may strenghten the role played by the mutated ARID1a allele in the impaired GR response with respect to other possible alterations (including the additional mutations identified by RNAseq) potentially present in this particular cancer cell line.

We agree with the reviewer that ideally we would over-express the full-length ARID1a in the SKOV3 cells and assess for changes in the RNAseq data. However several attempts have now been made to try to reintroduce the full length ARID1a construct back into the SKOV3 cells. Despite creating and obtaining a full length ARID1a construct complete with several tags and a Green fluorescent protein, transfecting the construct has proven incredibly difficult (See appendix A, B). Similarly trying to knock-down ARID1a in the SKOV3 cells has proved difficult (See appendix C). We now believe this to be impossible without further more invasive techniques such as electroporation, or methods outside our current expertise such as viral infection, both of which may also have many adverse effects on the cells and prevent us from properly deciphering the role of ARID1a. Therefore, we can only offer this comparison, which we have attempted to put into context appropriately in the manuscript.

2. Fig 1B, how do the Authors explain the partial retention of GR in the S100 fraction upon stimulation of SKOV3 cells ? Is ARID1a also expected to be involved in nuclear translocation of GR ? Should this be ARID1a-independent (see also point 1), this may suggest that glucocorticoid resistance of SKOV3 may have a heterogenous molecular basis.

The fractionation methodology used for these studies is known to yield a nuclear extract fraction containing the high salt extractable, chromatin bound , and a S100 fraction composed of low salt extractable, non chromatin associated nuclear proteins and cytoplasmic proteins (Yang et al., 1996, 1997; Liu et al., 1999a/b; Deroo et al., 2002; Tago et al., 2004). In our hands, ARID1a is only present in the nuclear (chromatin associated) fraction of the SKOV3 and HeLa cells and therefore it is not expected to be involved in the nuclear translocation of GR. However ARID1a may be involved in recruiting GR to the chromatin and this may explain why some GR is retained in the low salt S100 fraction. GR is not always depleted from the S100 fraction in every cell line in the presence of Dex, so this may also be a cell specific trait. In order to prove GR chromatin association is dependent upon ARID1a we would have to first be able to overexpress or knock-down the construct (see answer to question 1). We would like to thank the reviewer for this comment and have amended the text (line number 185-188) to further clarify this.

Figure 1. Data revealing the absence of ARID1a in our S100 fractions of both cell lines.

References: YANG, J. & DEFRANCO, D. B. 1996. Assessment of glucocorticoid receptor-heat shock protein 90 interactions in vivo during nucleocytoplasmic trafficking. Mol Endocrinol, 10, 3-13. YANG, J., LIU, J. & DEFRANCO, D. B. 1997. Subnuclear trafficking of glucocorticoid receptors in vitro: chromatin recycling and nuclear export. J Cell Biol, 137, 523-38. LIU, J. & DEFRANCO, D. B. 1999. Chromatin recycling of glucocorticoid receptors: implications for multiple roles of heat shock protein 90. Mol Endocrinol, 13, 355-65. LIU, J., XIAO, N. & DEFRANCO, D. B. 1999. Use of digitonin-permeabilized cells in studies of steroid receptor subnuclear trafficking. Methods, 19, 403-9. DEROO, B. J., RENTSCH, C., SAMPATH, S., YOUNG, J., DEFRANCO, D. B. & ARCHER, T. K. 2002. Proteasomal inhibition enhances glucocorticoid receptor transactivation and alters its subnuclear trafficking. Mol Cell Biol, 22, 4113-23. TAGO, K., TSUKAHARA, F., NARUSE, M., YOSHIOKA, T. & TAKANO, K. 2004. Regulation of nuclear retention of glucocorticoid receptor by nuclear Hsp90. Mol Cell Endocrinol, 213, 131-8.

Minor

3. Introduction, page 3: the last 15 lines or so, summarizing the results of the study, may be omitted or significantly shortened.

We would like to thank the reviewer for the feedback on this section that has now been significantly shortened.

4. Given the frequent occurrence of ARID1a mutations in cancer (perhaps also in patients not previously treated with glucocorticoids), do the Authors envisage that loss-of-function lesions, such as the one described in this study, may be only involved in secondary resistance to glucocorticoids or play additional roles in driving cancer formation (also regardless glucocorticoids exposure) ?

Again we would like to thank the reviewer for these questions. It is well known and largely published that ARID1a is a tumour suppressor. I have now included this into the introduction (line number 56-58).

Appendix A. Overexpression of a Full length ARID1a construct

A pcDNA6-ARID1a construct was obtained as bacteria in an agar stab (Addgene, UK). Bacteria were grown and plasmid DNA was extracted using a Qiagen maxi-prep kit (Qiagen, UK) following the manufacturer’s instructions. SKOV3 cells were grown for 24 hours in normal growth medium in 10cm2 plates to 70% confluence. The media was changed to a reduced serum Optimem (Gibco, Life Technologies, UK) an hour prior to transfection. A selection of quantity ratios of DNA to Lipofectamine2000 (Invitrogen, UK) were combined together and incubated for 5 minutes in Optimem at room temperature before being added to the cells (Table 1). Following an overnight incubation whole cell lysates were prepared for Western blotting. Anti-ARID1a antibodies (Bethyl Labs; A301-040A, A301-041A), a His-tag antibody (Cell signaling, 2365) and a loading control Beta Tubulin (Sigma, T4026) were used for protein detection.

Table 1: Transfection ratios of DNA to Lipofectamine2000. Transfecting pcDNA6-ARID1a into SKOV3 cells. Condition PcDNA6-ARID1a Lipofectamine2000 Optimem volume required (673.2 μg/ml) (μl) (μl) for 1ml (μl) Untransfected 0 30 970 Low DNA transfection 14 30 956 High DNA transfection 20 30 950 Very high DNA transfection 28 30 942

No change in the detection of ARID1a protein was observed (Figure 1). Beta tubulin is used as a loading control for SKOV3 whole cell lysates. Similarly, no His-tagged protein could be detected by Western Blot using an anti-His antibody in whole cell lysates or nuclear extracts in these cells (data not shown). As we had no positive control for the antibody we were therefore unable to determine whether the construct had been transfected into the cells.

Figure 1: Western blot showing no increase in the amount of ARID1a protein in whole cell lysates after attempted transfection of Addgene pcDNA6-ARID1a in SKOV3 cells. Beta Tubulin is used as an endogenous loading control.

B: Transfecting a full length ARID1a-GFPN1 construct

A fluorescent-tagged construct was considered to be beneficial in optimizing the transfection, in particular to aiding in the direct observation of the transfection efficiency during optimization. The following construct was created using a GFP-N1 backbone vector and incorporates a HIS-TAG (Figure 2). SKOV3 cells and HeLa cells (used as a comparison) were grown for 24 hours in normal growth medium in 10cm2 plates to 70% confluence. The media was replaced with reduced serum Optimem (Gibco, Life Technologies, UK). A variety of concentrations of DNA to Lipofectamine2000 (Invitrogen, UK) were used to determine the best transfection efficiency. ARID1a-GFPN1 DNA and Lipofectamine2000 were incubated together for 5 minutes in Optimem before addition to the cells. Following an overnight incubation cells were observed using a fluorescence microscope (Leica Biosystems, UK).

A.

B.

Figure 2: GFPN1 vector and ARID1a insert from Addgene vector. A) Represents the two constructs, GFPN1 and pcDNA-ARID1a, which are digested and ligated to create ARID1a- GFPN1. B) Schematic of ARID1a-GFPN1 and the restriction enzymes used to create construct.

However no fluorescence was observed using Lipofectamine2000 so other transfection reagents were used. K2 transfection reagent (Biontex Laboratories, UK) was the most successful transfection reagent used. However even using the ‘In-Cell 1000’ to visualize transfected cells, there was no tranfection of SKOV3 cells and very few cells transfected in HeLa cells (Figure 3). Consistent with this, Western blotting of SKOV3 cells again revealed a lack of increase in the ARID1a protein levels. As HeLa cells are regarded as relatively easy to transfect, the likelihood of overexpressing ARID1a in the SKOV3 cell line seemed highly unlikely without using a viral infection protocol, which we did not have access to.

Figure 3: ARID1a-GFPN1 transfection efficiency chart. Three full length ARID1a-GFPN1 clones (clone 1, 3 and 9) were transfected, using K2 transfection reagent, for 24 hours in HeLa cells. The majority of cells, indicated by the mean bar, do not show high cell fluorescence intensity compared to a GFPN1 transfected control and an untransfected (untrans) control (with only K2 transfection reagent). The mean fluorescence was very similar to that of untransfected cell group. Observation of cell images also revealed very few cells were fluorescent and suggested only a few individual cells were transfected.

C: Knock-down of ARID1a using shRNA with a pSUPER construct

Four sets of double stranded RNA oligos were designed to target ARID1a by short hairpin RNA (shRNA) knock-down when directionally cloned into a carrier vector, pSUPER (Table 2). Oligos contained a unique 19-nucleotide (nt) sequence derived from the ARID1a mRNA transcript, which was targeted for gene suppression. The forward and reverse oligos were designed to anneal and be cloned into the pSUPER vector, between the unique BglII and XHOI sites. This design enables the forward oligo to be positioned downstream of the pSUPER H1 promoter’s TATA box. The ARID1a gene sequence and therefore the 19 nt mRNA sequence corresponds to the sense strand of the subsequent pSUPER produced siRNA. The antisense strand of the siRNA binds to this mRNA region to promote cleavage of the siRNA. The forward oligo also contains a 9 nt spacer sequence which separates the 19 nt target sequence within both the sense and antisense strands. The oligo 5’ overhang corresponds to a BamHI site that is compatible with the plasmid’s 3’ BglII overhang therefore ligation destroys the BglII site producing a higher chance of obtaining positive clones. Cloning was successful and attempts were made to transfect DNA constructs into SKOV3 cells and HeLa cells.

Table 2: RNA oligos for shRNA ARID1a knock-down. Target ARID1a OLIGO Design Sequence cgtgtgtggagaacttaga 5’-GATCCCCcgtgtgtggagaacttagaTTCAAGAGAtctaagttctccacacacgTTTTTC-3’ (7268~)*, UTR 3’-GGGgcacacacctcttgaatctAAGTTCTCTagattcaagaggtgtgtgcAAAAAGAGCT-5’ (BamHI) (XHOI) ccaacaacatggcggacaa 5’-GATCCCCccaacaacatggcggacaaTTCAAGAGAttgtccgccatgttgttggTTTTTC-3’ (883~), ORF 3’-GGGggttgttgtaccgcctgttAAGTTCTCTaacaggcggtacaacaaccAAAAAGAGCT-5’ (BamHI) (XHOI) tggcagaaggaggagactt 5’-GATCCCCtggcagaaggaggagacttTTCAAGAGAaagtctcctccttctgccaTTTTTC-3’ (3834~)*, ORF 3’-GGGaccgtcttcctcctctgaaAAGTTCTCTttcagaggaggaagacggtAAAAAGAGCT-5’ (BamHI) (XHOI) gccaaggagagcagagtaa 5’-GATCCCCgccaaggagagcagagtaaTTCAAGAGAttactctgctctccttggcTTTTTC-3’ (2379~)*, ORF 3’-GGGcggttcctctcgtctcattAAGTTCTCTaatgagacgagaggaaccgAAAAAGAGCT-5’ (BamHI) (XHOI) Oligos were purchased from Sigma, UK

The construct DNA was incubated with Lipofectamine3000 (and Lipofectamine3000 additional reagent) (Invitrogen, UK) for exactly 5 minutes before addition to HeLa and SKOV3 cells. For a control, the PSUPER vector alone was transfected in the same way into the cells. Unfortunately, the efficiency of the transfection was low, as observed under a fluorescent microscope, and failed to knock-down the 250kDa ARID1a protein band seen on Western blots (Figure 4).

Figure 4: Western blot showing no knock-down in ARID1a protein levels following transfection with several pSUPER-ARID1a shRNA clones compared to the pSUPER vector alone. Lamin is used as an endogenous loading control for HeLa cell nuclear extracts.

*Manuscript Click here to view linked References

1 SKOV3 cells containing a truncated ARID1a protein 2 have a restricted genome-wide response to 3 glucocorticoids

4 F. E. Stubbsa, M. T. Birniea, S. C. Biddieb, S. L. Lightmana and B. L. Conway-Campbella

5

6 a) Henry Wellcome Laboratories for Integrative Neuroscience and Endocrinology, School of Clinical Sciences, 7 University of Bristol, Dorothy Hodgkin Building, Whitson Street, Bristol BS1 3NY, UK

8 b) Present Address: Simon

9 * Corresponding Author:

10 Dr Becky L. Conway-Campbell. Henry Wellcome Laboratories for Integrative Neuroscience and Endocrinology, 11 School of Clinical Sciences, University of Bristol, Dorothy Hodgkin Building, Whitson Street, Bristol BS1 3NY, UK. 12 +44 (0)117 3313138. Email [email protected]

13

14 List of Authors Contact Details:

15 Felicity E. Stubbs: [email protected]

16 Matthew T. Birnie: [email protected]

17 Simon C. Biddie: [email protected]

18 Stafford L. Lightman: [email protected]

19 Becky L. Conway-Campbell: [email protected]

20 Abstract

21 AT-rich interacting domain subunit 1a (ARID1a) is an essential SWI/SNF component frequently 22 mutated in human cancers. ARID1a mutations have also been associated with glucocorticoid 23 resistance, potentially related to the well-established role of the SWI/SNF complex in glucocorticoid 24 target gene regulation. Glucocorticoids are steroid hormones important for regulating many 25 physiological processes through the activation of the glucocorticoid receptor (GR). As GR interacts 26 directly with ARID1a, we hypothesized that a truncating ARID mutation would interfere with GR- 27 dependent gene regulation. Using high throughput RNA sequencing (RNA-SEQ) we show a restricted 28 glucocorticoid response in SKOV3 cells, which contain an inactivating ARID1a mutation. We also 29 show a lack of GR binding at the GR-dependent regulatory site in the Period 1 gene, which has 30 previously been shown to require chromatin remodelling. Taken together, our data suggests that 31 ARID1a may be required for regulation of a subset of glucocorticoid responsive genes. In the case of 32 SKOV3 cells, in which ARID1a is mutated, glucocorticoid-dependent transcriptional regulation of 33 these genes is significantly impaired.

34 Keywords 35 Glucocorticoids, Glucocorticoid Receptor, SWI/SNF, ARID1a, Transcription.

36 Abbreviations 37 GR – Glucocorticoid Receptor

38 GRE – Glucocorticoid response element

39 SWI/SNF - SWItch/Sucrose Non-Fermentable

40 BAF – BRG1 associated factor

41 ARID – AT-rich interaction domain

42 FKBP5 - FK506 binding protein 5

43 Per1 – Period 1

44 DUSP1 – Dual specificity phosphatase 1

45 ChIP – Chromatin Immunoprecipitation

46 FPKM - Fragments per kilobase per million mapped reads

47 mRNA – Messenger RNA

48

49 1. Introduction

50 Glucocorticoids are among the most commonly used pharmacological agents due to their potent 51 anti-inflammatory properties (1, 2). They are also extensively used as a cancer treatment due to 52 their ability to induce apoptosis and promote cell cycle arrest (3, 4, 5). However despite their 53 benefits, these drugs are often associated with several side effects (6, 7, 8, 9) and a significant 54 proportion of patients develop glucocorticoid resistance (10, 11). ARID1a mutations have been 55 linked to glucocorticoid resistance and are identified across a multitude of human cancers often 56 associated with poor patient prognosis (12). ARID proteins have previously been found to be tumour 57 suppressors and knockdown of their expression can promote cancer formation (13). ARID1a is an 58 essential component of the ATPase driven SWItch/Sucrose NonFermentable (SWI/SNF) chromatin- 59 remodelling complex. The protein’s C-terminal has previously been reported to directly interact with 60 GR (14, 15) (Supplementary figure 1). ARID1a interacts with AT-rich sequences within the DNA (16) 61 potentially recruiting GR to specific glucocorticoid response elements (GREs) to regulate gene 62 expression through dynamic mobilization of the chromatin. Determining the functional role of 63 ARID1a in GR signalling is therefore of great importance for understanding how mutations in the 64 SWI/SNF subunit could contribute to glucocorticoid resistance. 65 Glucocorticoids act through the binding of GRs located in the cell cytoplasm bound to chaperone 66 proteins (17, 18, 19). Upon ligand binding, GR translocates to the nucleus where it binds at GREs (20, 67 21, 22). Chromatin-remodelling by the SWI/SNF complex is a vital component of genomic GR 68 signalling, with chromatin being dynamically opened and closed at GREs in target genes regulating 69 the access of transcriptional machinery and RNA Polymerase II (23, 24). The SWI/SNF complex is 70 comprised of a single ATPase, important for driving the remodelling activities of the complex, and 71 several associated BRG-1 associated factors including ARID1a (BAF250a) (25). ARID1a could be an 72 important subunit of the SWI/SNF complex potentially facilitating and fine-tuning this GR mediated 73 transcriptional regulation for a subset of GR-dependent genes (26). Therefore absence of the 74 functional ARID1a protein may disrupt GR-dependent gene regulation and provide one possible 75 mechanism for development of glucocorticoid resistance. 76 SKOV3 cells are a human cell line that has been extensively used as a model to 77 study molecular mechanisms and outcomes of a mutation in ARID1a (27, 28, 29). SKOV3 cells contain 78 a truncating mutation of ARID1a in exon 3; a Glutamine (codon CAG) is altered to a premature stop 79 codon (TAG) at position 586 in the amino acid sequence (28). Such a truncating mutation results in 80 loss of the reported GR binding domain so may impact on GR-dependent gene regulation. To test 81 this, the glucocorticoid responsiveness of the SKOV3 cells has been assessed using RNA-SEQ 82 expression profiling over a 6 hour Dexamethasone (Dex) time course. As a technical control for our 83 analysis pipeline, we have conducted parallel assessment of the robustly glucocorticoid responsive 84 HeLa cell line (30, 31) known to contain a full-length functional ARID1a. 85 86 2. Methods 87 88 2.1 Cell culture 89 90 Human cervical cancer cell line, HeLa cells, and human ovarian adenocarcinoma, SKOV3 cells, were 91 obtained from the European Collection of Cell Cultures (ECACC; Sigma-Aldrich), each as a frozen 92 stock. Human HeLa cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM) with L- 93 glutamine and glucose (Gibco, Life Technologies, UK), 10% foetal calf serum (FCS, Invitrogen, UK) and 94 1% penicillin/streptomycin (Gibco, Life Technologies, UK). Human SKOV3 cells were maintained in 95 phenol red McCoy’s 5a modified medium (PAA, UK), 10% FCS (Invitrogen, UK) and 1% 96 penicillin/streptomycin (Gibco, Life Technologies, UK). Cells were cultured in 10cm2 plates (for RNA 97 extraction and RT-qPCR) OR 115cm2 flasks (for ChIP) and maintained in a humidified incubator (LEEC 98 Ltd., UK) at 37oC and 5% CO2 until 70% confluent. Cultures were washed 3 times in PBS 24 hours 99 prior to hormone addition and maintained in serum and phenol red free DMEM-F12 medium 100 supplemented with 100 μg/ml BSA (fraction V) (GIBCO) and 10 μg/ml transferrin (Sigma). A final 101 concentration of 100nM Dex was added to this culture medium.

102 2.2 Nuclear fractions

103 As previously described (32) the cells were resuspended in 300μl S1 buffer [10 mM pH 7.9 HEPES, 10 104 mM KCl, 1.5 mM MgCl2, 0.1 mM pH 8 EDTA] supplemented with 0.5 mM dithiothreitol (DTT), 50x 105 complete protease inhibitor (Roche, UK), 2mM NaF and 0.2mM NaVan. Samples were briefly spun at 106 2000 x g at 4 oC and the cytoplasmic S100 fraction was removed. The nuclear pellet was resuspended 107 in 1.2 volumes of cold S2 buffer [10 mM pH 7.9 HEPES, 400 mM NaCl, 1.5 mM MgCl2, 0.1 mM pH 8 108 EDTA, 5% glycerol] supplemented with 2mM NaF, 0.2 mM NaVan, 0.5mM DTT and Complete 109 protease inhibitor (Roche, UK). The pure nuclear fraction and cytoplasmic fraction were then stored 110 at -80 oC. A bicinchoninic acid assay (Pierce, Rockford, IL) was used to determine protein 111 concentrations.

112 2.3 Western blotting

113 As previously described in ref. Antibodies used were ARID1a A301-040A (Bethyl labs) and A301-041A 114 (Bethyl labs), GR E20 (Santa Cruz, sc-1003), Anti Beta Tubulin (Sigma, cat: T4026) and Anti Lamin (Cell 115 Signaling Technology, 2032). The membranes were washed and incubated with the appropriate 116 secondary antibody, ECL anti-rabbit, cat: NA934V, or ECL anti-mouse, cat: NA931V (GE Healthcare, 117 UK). Protein bands were detected using chemiluminescence; an enhanced chemiluminescence (ECL) 118 reagent (EZ-ECL Biological Industries, USA).

119 2.4 RNA extractions and Real time qPCR

120 RNA was purified using membrane columns (RNAeasy minikit, Qiagen, UK) following the 121 manufacturer’s guidelines. The RNA sample was diluted using nuclease free water and 1mg of RNA 122 was reversed transcribed into cDNA using a cloned Avian Myeloblastosis Virus (AMV) first strand 123 synthesis kit (Invitrogen, Life Technologies, UK) following the manufacturers guidelines. Real-time 124 PCR assays were performed using SYBR (Applied biosystems) and primers (see Supplementary table 125 1).

126 2.5 Chromatin Immunoprecipitation

127 After treatment, cells were fixed for 10 minutes in 1% formaldehyde, quenched with 0.125M 128 glycine and washed with PBS. Nuclear extracts were prepared as described in 2.2, then resuspended 129 in 6 Vol MNase digestion buffer [50mM Tris-HCl, pH 7.5, 4mM MgCl2, 1mM CaCl2, 0.32mM sucrose, 130 2mM NaF, 0.2mM NaVan] with 2 units of MNase (cat: N5386, Sigma, UK) per 180μg chromatin for 15 131 minutes at 37 oC. MNase digestion was stopped with addition of 5mM EDTA pH8 on ice. The nuclear 132 pellet was resuspended in 300μl sodium dodecyl sulfate (SDS) lysis buffer [1% SDS, 10mM EDTA, 133 50mM Tris-HCl pH 8.1], supplemented with complete protease inhibitor (Roche Diagnostics, Burgess 134 Hill, UK), 2mM NaF and 0.2mM NaVan. Lysates were cleared of debris by centrifugation at 16200 x g 135 for 15 minutes at 4oC and the soluble chromatin was collected for chromatin immunoprecipitation 136 (ChIP). For each ChIP reaction, 30μg samples were processed with H300 GR antibody (Santa Cruz sc- 137 8992) or rabbit non-immune serum (Santa Cruz sc-2027). RT-qPCR was performed using SYBR 138 reagents (Applied Biosystems) and the following Per1 primers.

139 Primer sequences (h, human; 5’-3’)

140 Distal hPER1, F ACAGGACGGCTGTCGTTTTG 141 Distal hPER1, R CGCACTTGGGAACATCATGT 142 143 Distal hPER1, F TCATGTTCTCTTGGCTGGTG (33) 144 Distal hPER1, R GGCCCCCTTCCTACTAATCC (33) 145 146 Proximal hPER1, F CTAGTCCGAAGTGGGCTGAC (33) 147 Proximal hPER1, R CCGGTCTTCTTGCTCGTTAC (33) 148 149 Exon 19 hPER1, F GCCTTGGTGCTCCCTAACTA 150 Exon 19 hPER1, R TCTGGAGTGCCCCATAAGGA 151 152 2.6 High throughput RNA Sequencing 153 154 High throughput RNA sequencing was performed (Bristol genomics facility) and the raw data was 155 analysed using Galaxy (www.galaxyproject.org). Three lanes were uploaded for each RNA sample 156 and for each condition an n of at least 3 were assessed. The Cuff Diff parameters included were 157 geometric library organization, pooled dispersion estimation, 0.05 False discovery rate, Min 158 alignment count 10, multi-read correct, bias correction and cuff-links effective length correction. 159 Differential gene expression was calculated as fragments per kilobase per million mapped reads 160 (FPKM) values; summed fragments of each transcript with same gene ID. After Benjamini-Hochberg 161 false discovery correction, genes with adjusted p values less than 0.05 were considered as 162 differentially expressed genes. 163 164 3. Results

165 3.1 Decreased ARID1a protein expression in SKOV3 cells

166 We have confirmed a mutation (C1756T) in exon 3 of ARID1a (Supplementary Fig 2) resulting in a 167 premature stop codon (Q586*) in SKOV3 cells as described previously (28). Interestingly, we can 168 detect both transcripts in the sequencing data, consistent with a heterozygous mutation. Western 169 blotting was used to assess whether this mutation resulted in loss of ARID1a protein (Figure 1). A 170 strong 250kDa band was detected in the nuclear fraction of the positive control HeLa cells (Figure 171 1A) but not detected in HeLa cells after ARID1a siRNA knockdown (Figure 1B) confirming the band to 172 be ARID1a. As shown in Figure 1A a 250 kDa band was also detected in the nuclear fraction of SKOV3 173 cells indicating the presence of full-length ARID1a. Using Image J Densitometry 174 (http://imagej.nih.gov/ij) SKOV3 cells were found to express approximately 62 % of the full-length 175 ARID1a protein expressed in HeLa cells. Taken together, the RNA-SEQ and Western blot results 176 indicate that the ARID1a mutation in SKOV3 cells is heterozygous. ARID1a has been suggested to be 177 a haplosufficient tumour suppressor with a large proportion of cancers containing a heterozygous 178 ARID1a mutation (27, 34).

179 Finally and importantly for this study, both cell lines possess a full-length, functional GR indicated by 180 a band detected at the expected size of 97kDa, which was detected in the nuclear fraction only after 181 Dex treatment (Figure 1C, D). Similar GR protein levels were found for both cell lines in the absence 182 of Dex. In contrast to the near complete depletion of GR from the S100 fraction after Dex treatment 183 in HeLa cells, a visible GR band remains in the S100 fraction after Dex treatment in SKOV3 cells. This 184 is highly suggestive of the presence of ‘a pool of inactive GR’ remaining after Dex treatment in the 185 SKOV3 cells. This reflects an impairment of GR chromatin interactions or stability in SKOV3 cells 186 potentially due to the loss of some full length ARID1a. ARID1a was not found in the S100 low salt 187 cytoplasmic fraction for either cell line (data not shown). This is in contrary to previous suggestions 188 that ARID1a is shuttled from the cytoplasm to the nucleus (35). As ARID1a appears to be associated 189 with the chromatin it thereby would not be essential for GR translocation from the cytoplasm. 190 Despite this, there was still a robust activation response detected in the SKOV3 cells, with a 191 significant translocation of GR to the ‘active nuclear fraction’.

192

193

194

195

196

197

198

199

200

201 Figure 1: Western blots indicating the presence and subcellular localization of ARID1a and GR in HeLa cells and SKOV3 202 cells with or without Dex treatment. A) ARID1a protein is observed in the nuclear fraction of both cell lines. Lamin is used 203 as an endogenous loading control specific to the nuclear extract fraction (NE). B) ARID1a is observed to be knocked down 204 in HeLa cells using reverse transfections of a combination of four siRNAs for 48 hours. C) Western blot showing GR in HeLa 205 cells in the nuclear extract (NE) and cytoplasm (S100) with or without Dex. The translocation of GR into the nuclear fraction 206 upon Dex treatment indicates a functional GR protein. Beta Tubulin is used as a control specific to the cytoplasm. D) 207 Western blot showing GR in SKOV3 cells with or without Dex in both the NE and cytoplasm (S100). An increase in GR into 208 the nucleus upon Dex treatment indicates that GR is functional. 209 210

211 3.2 Basal expression profiling characterises cell-specific differences between SKOV3 and 212 HeLa cells

213 It was first important to determine cell-specific basal expression differences between the SKOV3 214 and HeLa cells prior to assessing glucocorticoid responsiveness. Therefore, RNA-SEQ expression 215 profiles of the two cell lines were compared in the absence of glucocorticoid (time 0hr). As expected 216 there are several cell-specific expression differences at baseline (pre-treatment, time 0hr) as shown 217 by the heat map, Figure 2A. By comparing these basal groups for both cell lines, 3171 genes were 218 identified as being more highly expressed in HeLa cells compared to SKOV3 cells. Additionally, 3507 219 genes are more highly expressed in the SKOV3 cells compared to the HeLa cells (Supplementary 220 table 1). This is the expression profile that defines each cell line, indicating expected cell-specific 221 differences in their respective transcriptomes. Despite the difference in protein observed from 222 Western blotting, basal ARID1a mRNA levels are very similar in both cell lines. Basal GR mRNA levels 223 in SKOV3 cells are approximately 3 times lower than the levels in HeLa cells. Importantly, there are 224 large numbers of genes that are common to both cell lines (Figure 2B, C). Figure 2C shows a high 225 correlation between basal expression values of genes with FPKM values greater than 1 for both HeLa 226 and SKOV3 cells with a Pearson correlation coefficient of 0.86. Many commonly expressed 227 glucocorticoid responsive genes such as GILZ (38), IRS2 and THBD (39) have similar basal expression 228 levels. However, some largely ubiquitous genes such as Per1 and DUSP1 are expressed at lower 229 levels in the SKOV3 cells; the basal mRNA levels are approximately 3 times lower for Per1 and 2 230 times lower for DUSP1 compared to HeLas. Some widely expressed genes such as DKK1 (40) have 231 much higher basal expression in the SKOV3 than in the HeLas; basal expression is approximately 14 232 times higher in SKOV3 cells. However, it is predominantly the overall pattern of global changes after 233 Dex treatment that is important for a comparison of the glucocorticoid responsiveness of these two 234 cell lines.

235

236 237 Figure 2: Basal gene expression and genes regulated by Dex in HeLa versus SKOV3 cells. A) Heat map indicating the Log10 238 expression of genes after 0hr (control), 1hr, 3hr and 6hr Dex treatment in both HeLa and SKOV3 cells (includes all genes 239 with a value of >1 FPKM at basal (0hr) gene expression for either cell line). Blue and Red represent lowest and highest 240 expression levels respectively, with intermediate expression levels represented by the colour gradient indicated in the key. 241 B) A Venn diagram to show the total number of significantly regulated Dex responsive genes, in HeLa cells compared to 242 SKOV3 cells, that are regulated at any stage during the 1, 3 and 6 hour timecourse as determined by differential expression 243 analysis of RNA-SEQ data as described by Trapnell and colleagues (36, 37). All genes shown are significantly differentially 244 regulated at the P < 0.05 level and FPKM values > 1. C) Pearson correlation coefficient (r) = 0.86 and P value (p) = <0.001 245 between genes with FPKM values > 1 at basal levels for HeLa and SKOV3 cell (in Log2). Data in red represents genes with 246 >1.5x fold expression in SKOV3 compared to HeLa cells. Data in blue represents genes with >1.5x fold expression in HeLa 247 compared to SKOV3 cells. A line of best fit is also shown. D, E) Volcano plot showing differentially expressed genes (of 248 genes with basal levels FPKM values > 1) in HeLa cells (D) and SKOV3 cells (E) following 3 hours of Dex treatment. The 249 negative Log10 transformed P values test the null hypothesis of no expression level change between basal level controls 250 and cells with 3 hours of 100nM Dex treatment (y-axis) and are plotted against the average Log2 fold changes in expression 251 (x-axis). Data for genes that were not differentially expressed are plotted in black. Data for genes that were differentially 252 expressed (as determined by Benjamini-Hochberg correction analysis with FDR = 0.05 for RNA-SEQ) are plotted in blue (P < 253 0.05), green (P < 0.01) and red (P < 0.001). F, G, H, I) Venn diagram analysis shows overall numbers of regulated genes over 254 the timecourse in each cell line. Significant gene induction (F, G) and repression (H, I) in HeLa (F, H) and SKOV3 cells (G, I) 255 determined by differential expression analysis of RNA-SEQ data (36, 37; using Benjamini-Hochberg correction FDR=0.05), 256 following Dex treatment of 1, 3 and 6 hours compared to 0 hour control. All genes shown are significantly differentially 257 regulated at the P < 0.05 level and FPKM values > 1.

258

259 3.3 SKOV3 Cells Exhibit a Restricted Genome-Wide transcriptional response to 260 Glucocorticoids

261 Visualisation of global changes in gene expression over the timecourse of Dex treatment (Figure 2A) 262 revealed striking differences between the two cell lines, which is further shown quantitatively in the 263 Venn Diagram analysis of Dex-regulated genes (Figure 2B). Figure 2D and 2E compares the 264 glucocorticoid responsiveness of genes common to both cell lines and demonstrates greater fold 265 change, seen as both higher inductions (Log2 fold change > 0) and greater repressions (Log2 fold 266 change < 0) of a larger number of genes in HeLa cells (Figure 2D) compared to SKOV3 cells (Figure 267 2E). The total number of induced or repressed genes, over the full timecourse, for both cell lines is 268 shown in by Venn diagram (Figure 2F, G, H, I). Over the full Dex timecourse, a total of 101 269 transcripts were upregulated in SKOV3 cells compared to 570 in HeLa cells (Figure 2 F, G). A total of 270 49 transcripts were downregulated by Dex in the SKOV3 cells compared to 350 in the HeLa cells 271 (Figure 2 H, I).

272 Real-time quantitative polymerase chain reaction (qPCR) was used to validate differential regulation 273 of a selection of targets identified in the RNA-SEQ analysis. Differential regulation of the well-known 274 Dex-regulated genes Per1, FKBP5, BIRC3, DUSP1, SLIT2, FR-alpha and SLIT3 is shown in Figures 3B, 275 and 3C. Both cell lines had a significant induction of Per1 by 6 hours of Dex treatment, although 276 induction was far more rapid in the HeLa cells reaching significant upregulation at 1hour (Figure 3A). 277 DUSP1 was induced by 1 hour in HeLa cells but only showed a non-significant trend towards 278 induction by 6 hours of Dex treatment in the SKOV3 cells. BIRC3 and FKBP5 were significantly 279 induced in HeLa cells (Figure 3A) but not in SKOV3 cells, which is comparable to the RNA-SEQ data. 280 FR-alpha, which is highly Dex inducible in some cancer cell lines, is not regulated by Dex in SKOV3 281 cells (Figure 3B) (41). SLIT2 and SLIT3 were not regulated by GR in SKOV3 cells, despite being 282 repressed by glucocorticoids in the majority of ovarian cell lines (31) (Figure 3B). All these results 283 were consistent with the RNA-SEQ differential expression data and provided supporting evidence for 284 impaired GR-dependent gene regulation in SKOV3 cells. Overall RT-qPCR analysis confirmed the 285 restricted transcriptional response shown by the sequencing data.

286 3.4 SKOV3 Cells Show a Delayed Transcriptional Response to Dexamethasone

287 RNA-SEQ analysis revealed fewer genes to be responsive to Dex following 1 hour of treatment in 288 the SKOV3 cells compared to the HeLa cells. Only 11 genes were induced and 7 genes repressed 289 after 1 hour of Dex treatment in SKOV3 cells. In comparison, 94 genes were induced and 17 290 repressed after 1 hour Dex treatment in HeLa cells (Figure 2 F, G). In HeLa cells, 72 genes (12.6% of 291 all Dex inducible genes in HeLa cells) are induced throughout the Dex timecourse. A further 182 292 genes (31.9% of all Dex inducible genes in HeLa cells) were induced at both the 3 hour and 6 hour 293 time points. Interestingly, in contrast to HeLa cells no genes are induced throughout the entire 294 timecourse in SKOV3 cells. However, 21 genes were continually induced from the 3 hour timepoint 295 (only 20.8% of all SKOV3 Dex inducible genes). Furthermore, some of the genes regulated in 296 common between the two cells types exhibited a delayed induction in the SKOV3 compared to the 297 Hela cells. For Per1, gene induction was more gradual and progressed at each time interval in the 298 SKOV3 cells in comparison to the control HeLa cells where induction reached a significant increase at 299 the earlier 1 hour time point (Figure 3 A). A few other highly Dex inducible genes, DUSP1, GILZ, 300 CEBPD, NFKBIA, DKK1 and ZFP36, similarly to Per1, are induced from 1 hour of Dex treatment in 301 HeLa cells but induction in SKOV3 cells is only observed at 3 hours of Dex treatment. These genes are 302 positioned in the central area of the Venn diagrams in Figure 3C and 3D showing genes induced by 3 303 hours and 6 hours of Dex treatment in SKOV3 cells respectively, compared to genes induced from 1 304 hour of Dex treatment in HeLa cells. These two diagrams indicate the genes that are regulated in 305 both cell lines from 3 hours of Dex treatment. Despite a delay in the transcriptional response of Dex 306 inducible genes in SKOV3 cells there is no observed delay in Dex repressed genes. Previous studies 307 have shown that the majority of downregulated genes are repressed less rapidly than the majority of 308 upregulated genes are induced (42, 43). Therefore we cannot rule out that repression may occur at a 309 later timepoint. Despite this, consistent with a delay in repression fewer genes are repressed in 310 SKOV3 cells, 49 genes, in comparison to HeLa cells, 350 genes over the 6 hour timecourse. 311 Conversely, in contrast to induced genes the vast majority of repressed genes have previously been 312 shown to remain unaffected by the ablation of the SWI/SNF function (26). Overall this apparent loss 313 of ‘rapid’ regulation of several genes in SKOV3 cells compared to HeLa cells and may suggest either a 314 delay in GR binding, or a delay in the recruitment of required transcriptional components necessary 315 for GR-dependent gene regulation. This could include binding of co-regulators such as the SWI/SNF 316 complex or any factors of the transcriptional preinitiation complex including RNA Polymerase II.

317

318 3.5 Loss of Prolonged Glucocorticoid gene regulation in SKOV3 cells

319 RNA-SEQ data revealed loss of induction of a large number of glucocorticoid responsive genes by 6 320 hours of Dex treatment in SKOV3 cells. This is in stark contrast to HeLa cells, where a large 321 proportion of genes showed prolonged induction that remained upregulated at 6 hours (Figure 2 F, 322 G). Only 20.8% of genes remained significantly upregulated from 3 hours of Dex treatment in SKOV3 323 cells compared to 44.6% of Dex inducible genes in HeLa cells. In SKOV3 cells, 90.9% of genes (10/11) 324 induced following 1 hour of Dex treatment returned to baseline expression levels throughout the 325 timecourse. Additionally 62.5% of genes (35/56) induced following 3 hours of Dex treatment showed 326 loss of induction by 6 hours of treatment. In contrast, in HeLa cells, only 13.8% of genes (13/94) 327 induced by 1 hour of Dex treatment showed loss of induction across the timecourse. Only 30% of 328 genes (109/393) induced following 3 hours of Dex treatment returned to basal levels by 6 hours of 329 treatment. In SKOV3 cells, 59% of Dex repressed genes do not remain repressed by 6 hours of Dex 330 treatment. This is a large proportion of Dex repressed genes when in comparison 87.7% of repressed 331 genes in HeLa cells remain downregulated at 6 hours of Dex treatment. The loss of prolonged Dex 332 regulation of genes in SKOV3 cells throughout the timecourse may indicate a loss of GR-dependent 333 gene regulation caused by the loss of ARID1a protein 334 A.

B.

C. D.

335 336 337 338 Figure 3. Pattern of gene induction for common glucocorticoid responsive genes in SKOV3 cells compared to HeLa cells. 339 A - B) Fold change in mRNA expression normalised to β glucuronidase for SKOV3 cells (Purple) and HeLa cells (Blue) during 340 a Dex treatment timecourse. Data is represented at mean +/- Standard Error Mean (SEM). A) In both HeLa and SKOV3 cells 341 the mRNA expression of Per1 is significantly increased following 6 hours of Dex treatment. In HeLa cells the mRNA 342 expression of Per1 is also significantly increased at times 0, 1 and 3 hours. DUSP1 is significantly induced in HeLa cells at all 343 timepoints of Dex treatment but only shows a trend towards induction over the timecourse in SKOV3 cells. Induction of 344 BIRC3 and FKBP5 are also shown in HeLa cells but not in SKOV3 cells. For B) Two-way ANOVA with Dunnett’s post-hoc test 345 with all comparisons to the time 0 control were used and statistical significance indicated at P < 0.05 level (*), P < 0.01 (**), 346 P < 0.001 (***), P < 0.0001 (****). C) Lack of regulation of commonly reported, ovarian specific, Dex regulated genes in 347 SKOV3 cells FR-alpha, SLIT2 and SLIT3. Data is represented at mean +/- SEM. One-way ANOVA with Dunnett’s post-hoc test 348 was used, with all comparisons to the time 0 hour control. No statistically significant differences were detected. C) Venn 349 diagram analysis shows number of genes induced by 3 hour Dex treatment in SKOV3 cells compared to genes induced by 350 Dex in HeLa cells at 1, 3 and 6 hours. D) Genes induced by 6 hour Dex treatment in SKOV3 cells compared to genes induced 351 by Dex in HeLa cells at 1, 3 and 6 hours. For C) and D) all genes shown are significantly differentially regulated at the P < 352 0.05 level following Benjamini-Hochberg correction and FPKM values > 1. 353 354 355 356 357 358 Figure 4: GR binding at the Per1 proximal and distal GREs determined by ChIP studies. A) Schematic of the human Per1 359 gene and the position of the distal and proximal GRE sites. B) GR binding at the Distal and Proximal GREs in HeLa cells. C) 360 GR binding at the Distal and Proximal GREs in SKOV3 cells. One way ANOVA with Dunnett’s Post-hoc Test *** P<0.001. 361 362 363 364 3.6 SKOV3 cells have Decreased GR binding at a Chromatin Remodelling inducible GRE in 365 Per1 366 367 Per1 was chosen as a candidate gene to assess GR binding. Per1 contains two GR binding sites, the 368 first more distal site is a glucocorticoid and DNase1 hypersensitive GRE whereas the second proximal 369 site is an inducible GRE site requiring chromatin remodelling (Figure 4 A) (44). Therefore, it provided 370 a model for observing GR binding at a pre-accessible GR binding region, and at a site that is thought 371 to require GR to mediate further chromatin remodelling potentially through the recruitment of the 372 SWI/SNF complex. As both qPCR and RNA-SEQ suggested a delay in Per1 gene induction in response 373 to Dex and an overall lower level of Per1 expression and induction in SKOV3 cells compared to HeLa 374 cells it was important to determine whether these changes were due to the identified ARID1a 375 mutation. 376 GR binding at the Per1 distal GRE region increased significantly following a 30 minute treatment 377 with Dex in both cell lines compared to untreated controls (Figure 4 B, C). It should be noted that GR 378 binding at the distal GRE was lower in the SKOV3 cells compared to that in the HeLa cells. At the 379 proximal Per1 GRE region (within intron 1) GR binding increased following 30 minutes Dex treatment 380 in HeLa cells but not in SKOV3 cells (Figure 4 B, C). This loss of GR binding at the proximal site may 381 suggest decreased chromatin accessibility at this site. This could suggest that ARID1a is important for 382 promoting the accessibility of this site for GR binding. This is potentially a mechanism for pre-setting 383 the chromatin architecture prior to glucocorticoid treatment by primary chromatin remodelling. 384 Alternatively this proximal GRE may not be accessible for GR binding in SKOV3 cells and may be pre- 385 set by other mechanisms to be inaccessible to GR binding. This result becomes difficult to interpret 386 with confidence in light of the lower GR binding result at the distal site in the SKOV3 cells relative to 387 the HeLa cells. 388 389 3.7 Other gene mutations and loss of RNA expression detected in human SKOV3 cells 390 391 In addition to confirming the expected C to T mutation, which alters a Glutamine (codon CAG) to a 392 premature stop codon (TAG) at position 586 in the amino acid sequence, identified in 32% of 393 transcripts in SKOV3 cells (Supplementary Figure 2). Our RNA-SEQ data also confirmed a truncating 394 mutation in exon 24 of BAF155 (SMARCC1) (45) (Supplementary Figure 3) and a mutation in PIK3CA 395 (28) in which a Histidine (CAT) is changed to an Arginine (CGT) in 45% of transcripts (Supplementary 396 Figure 4). Another important interaction of the ARID1a protein is the direct binding of the protein to 397 the cell cycle regulator (46). However RNA-SEQ data revealed negligible P53 expression in 398 SKOV3 cells, which is in contrast to HeLa cells. 399 400 4. Discussion 401 402 Previously luciferase reporter assays have been used to suggest ARID1a is important for GR 403 activity (15). Here we provide evidence of a cell line with an ARID1a mutation with a restricted 404 glucocorticoid response. Notably, SKOV3 cells possess a largely restricted GR-dependent 405 transcriptional response following Dex treatment when compared to HeLa cells. There is also an 406 apparent delay in transcriptional regulation as well as a marked loss of transcriptional regulation of 407 GR-dependent genes over a 6 hour Dex timecourse in SKOV3 cells, which contrasts the dynamic and 408 robust transcriptional response observed in HeLa cells. ChIP assays assessing GR binding at 409 regulatory sites of the Per1 gene reveals a loss of GR binding at the proximal inducible GRE 410 containing site in the SKOV3 cells despite significant binding at this site in HeLa cells. This could 411 suggest that ARID1a is required to pre-set the accessibility of this GRE for GR binding. Although there 412 are significant differences between the RNA transcriptional responses and GR binding to Dex 413 treatment between these two cells, which could indicate differences due to the presence of a 414 mutated ARID1a, these may also be due to inherent differences in cell type and related to differing 415 cellular functions. 416 417 Multiple mutations identified in SKOV3 cells 418 419 Notably, the SKOV3 cell line has been widely used to study ARID1a (27-29), however the findings 420 emerging from our genome-wide study now reveal these cells to be a less than ideal model for this 421 purpose. In fact, several other mutations and loss of proteins have been observed in this cell line. It 422 is also largely recognised that cell lines containing an ARID1a mutation do also co-occur with other 423 gene mutations. Therefore loss of ARID1a alone may not result in such large-scale loss of 424 glucocorticoid responsiveness and may be a result of several mutations. Recent studies have shown 425 that the addition of a PIK3CA mutation as well as an ARID1a mutation can help promote tumour 426 formation in ovarian cancer cells (47). PIK3CA encodes the catalytic subunit of PI-3K and mutations 427 that co-occur with ARID1a often lead to increased PI-3K, and subsequently AKT, activity (47-49). PI- 428 3K is also known to interact with GR (50). Therefore the PIK3CA mutation in SKOV3 cells may 429 interfere with PI-3K-GR interactions and could be a reason for altered GR mediated gene regulation 430 in this cell line. BAF 155, another subunit of the SWI/SNF complex, has also been shown to have a 431 truncating mutation (51). Although BAF57 and ARID1a (BAF 250) have been predicted to be the main 432 mediators of an interaction of GR with the SWI/SNF complex using Protein interactions by structural 433 matching (PRISM) techniques (52), other BAF subunits may also be important for this interaction. 434 BAF155 was shown to be critical for subunit associations within the SWI/SNF complex (52), therefore 435 a mutation in BAF155 may alter interactions between other subunits such as BAF57 and ARID1a 436 within the complex. Interestingly deletion of the C-terminal of BAF155 has been shown to disrupt 437 the ability of this subunit to regulate the stability of BAF57 (51, 52). Therefore the restrictive GR 438 gene response to Dex treatment in SKOV3 cells may be due to loss of BAF57 from the SWI/SNF 439 complex, which is important for an interaction of the complex with GR. P53 is another important 440 factor known to directly interact with ARID1a (46) and it is now widely recognized that SKOV3 cells 441 are P53-negative, which again is supported by our RNA-SEQ data. GR-dependent cell cycle arrest can 442 often require P53 (53-55) and loss of P53 may impact upon regulation of a specific subset of GR- 443 dependent genes. Therefore the differences in GR-dependent gene regulation observed may result 444 from any combination of these mutations. Further studies assessing the impact of loss of other 445 factors alongside loss of ARID1a may therefore be useful in assessing whether an impact on GR gene 446 regulation occurs through a combination of mutations that frequently occur following ARID1a 447 mutations observed in cancer. 448 449 Truncated ARID1a protein 450 451 Another factor which could contribute to the loss of glucocorticoid responsiveness is the potential 452 presence of small truncated ARID1a protein of the first 585 amino acids of the N-terminal that may 453 still be present in SKOV3 cells alongside the full-length ARID1a protein. There is a lack of evidence on 454 the role of the ARID1a-NTD but previous studies have suggested it is required for GR activation and 455 overexpression of the ARID1a-CTD using luciferase reporter assays results in loss of GR activity (15). 456 An LXXLL motif has however been identified in the ARID1a-NTD (295-299 amino acids) (56), which 457 may be important in binding coactivators of GR regulation. One previous study overexpressed LXXLL 458 containing peptides that interact with ER (57). They showed that if these peptides mimic the 459 interactions between the receptor and endogenous co-factors they can function in a dominant 460 negative manner to disrupt interactions and prevent transcription (57). Therefore, the 461 overexpression of such a small truncated ARID1a N-terminal protein containing an LXXLL may act as 462 a dominant negative and interfere with the function of GR or other NRs that require interactions 463 with specific cofactors. The presence of a small N-terminal LXXLL containing inactive ARID1a protein 464 could potentially bind to and sequester certain co-factors thereby interfering with GR-dependent 465 gene regulation in SKOV3 cells. Future investigation may be aimed at further determining whether 466 the ARID1a-NTD is translated and its role in GR-dependent gene regulation by studying the effect of 467 overexpression of this domain in an ovarian cell line without an ARID1a mutation. 468 469 Summary 470 471 Despite the many caveats of the SKOV3 cell model, our genome-wide study has revealed a globally 472 restricted transcriptional profile of GR regulated genes in these cells. This finding has potentially 473 highlighted a functional role for ARID1a and the SWI/SNF complex in modifying the chromatin 474 landscape for regulation of large numbers of GR regulated genes. Importantly, our study has lead to 475 a potential mechanism to further explore in the clinical problem of glucocorticoid resistance, as 476 ARID1a mutations have also been linked to glucocorticoid resistance in Leukaemia patients (12). 477 Therefore further studies into ARID1a could subsequently identify subsets of cancer patients, which 478 are likely to develop glucocorticoid resistance and could lead to identification of new therapeutic 479 targets for these patients. 480 481 Acknowledgements 482 483 All materials and consumables were provided by funding from Wellcome Trust Grant 089647/Z/09/Z. F.E. 484 Stubbs was supported by a scholarship provided by the Needham Cooper Charitable Trust. 485 486 487 488 1. BEATO, M. & KLUG, J. 2000. Steroid hormone receptors: an update. Hum Reprod Update, 6, 489 225-36. 490 2. MCMASTER, A. & RAY, D. W. 2008. Drug insight: selective agonists and antagonists of the 491 glucocorticoid receptor. Nat Clin Pract Endocrinol Metab, 4, 91-101 492 3. PLANEY, S. L. & LITWACK, G. 2000. Glucocorticoid-induced apoptosis in lymphocytes. 493 Biochem Biophys Res Commun, 279, 307-12. 494 4. RUTZ, H. P. 2002. Effects of corticosteroid use on treatment of solid tumours. Lancet, 360, 495 1969-70. 496 5. HERR, I. & PFITZENMAIER, J. 2006. Glucocorticoid use in prostate cancer and other solid 497 tumours: implications for effectiveness of cytotoxic treatment and metastases. Lancet Oncol, 498 7, 425-30. 499 6. OLEFSKY, J. M. & KIMMERLING, G. 1976. Effects of glucocorticoids on carbohydrate 500 metabolism. Am J Med Sci, 271, 202-10. 501 7. WALKER, B. R. 1996. Abnormal glucocorticoid activity in subjects with risk factors for 502 cardiovascular disease. Endocr Res, 22, 701-8. 503 8. KERSHNER, P. & WANG-CHENG, R. 1989. Psychiatric side effects of steroid therapy. 504 Psychosomatics, 30, 135-9. 505 9. PIPER, J. M., RAY, W. A., DAUGHERTY, J. R. & GRIFFIN, M. R. 1991. Corticosteroid use and 506 peptic ulcer disease: role of nonsteroidal anti-inflammatory drugs. Ann Intern Med, 114, 507 735-40. 508 10. SHER, E. R., LEUNG, D. Y., SURS, W., KAM, J. C., ZIEG, G., KAMADA, A. K. & SZEFLER, S. J. 509 (1994) Steroid-resistant asthma. Cellular mechanisms contributing to inadequate response 510 to glucocorticoid therapy. J Clin Invest, 93, 33-9. 511 11. LEUNG, D. Y. & BLOOM, J. W. (2003) Update on glucocorticoid action and resistance. J 512 Allergy Clin Immunol, 111, 3-22. 513 12. POTTIER, N., YANG, W., ASSEM, M., PANETTA, J. C., PEI, D., PAUGH, S. W., CHENG, C., DEN 514 BOER, M. L., RELLING, M. V., PIETERS, R., EVANS, W. E. & CHEOK, M. H. 2008. The SWI/SNF 515 chromatin-remodeling complex and glucocorticoid resistance in acute lymphoblastic 516 leukemia. J Natl Cancer Inst, 100, 1792-803. 517 13. WU, R., WANG, T. & SHIH, L. 2014. The emerging roles of ARID1a in tumor suppression. 518 Cancer Biol Ther, 15, 655-664 519 14. NIE, Z., XUE, Y., YANG, D., ZHOU, S., DEROO, B. J., ARCHER, T. K. & WANG, W. 2000. A 520 specificity and targeting subunit of a human SWI/SNF family-related chromatin-remodeling 521 complex. Mol Cell Biol, 20, 8879-88. 522 15. INOUE, H., FURUKAWA, T., GIANNAKOPOULOS, S., ZHOU, S., KING, D. S. & TANESE, N. 2002. 523 Largest subunits of the human SWI/SNF chromatin-remodeling complex promote 524 transcriptional activation by steroid hormone receptors. J Biol Chem, 277, 41674-85. 525 16. CHANDLER, R. L., BRENNAN, J., SCHISLER, J. C., SERBER, D., PATTERSON, C. & MAGNUSON, T. 526 2013. ARID1a-DNA interactions are required for promoter occupancy by SWI/SNF. Mol Cell 527 Biol, 33, 265-80. 528 17. DITTMAR, K. D., DEMADY, D. R., STANCATO, L. F., KRISHNA, P. & PRATT, W. B. 1997. Folding 529 of the glucocorticoid receptor by the heat shock protein (hsp) 90-based chaperone 530 machinery. The role of p23 is to stabilize receptor.hsp90 heterocomplexes formed by 531 hsp90.p60.hsp70. J Biol Chem, 272, 21213-20 532 18. PRATT, W. B., SILVERSTEIN, A. M. & GALIGNIANA, M. D. 1999. A model for the cytoplasmic 533 trafficking of signalling proteins involving the hsp90-binding immunophilins and p50cdc37. 534 Cell Signal, 11, 839-51. 535 19. VANDEVYVER, S., DEJAGER, L. & LIBERT, C. 2012. On the trail of the glucocorticoid receptor: 536 into the nucleus and back. Traffic, 13, 364-74 537 20. SAVORY, J. G., HSU, B., LAQUIAN, I. R., GIFFIN, W., REICH, T., HACHÉ, R. J. & LEFEBVRE, Y. A. 538 1999. Discrimination between NL1- and NL2-mediated nuclear localization of the 539 glucocorticoid receptor. Mol Cell Biol, 19, 1025-37 540 21. YAMAMOTO, T., NAKAHATA, Y., TANAKA, M., YOSHIDA, M., SOMA, H., SHINOHARA, K., 541 YASUDA, A., MAMINE, T. & TAKUMI, T. 2005. Acute physical stress elevates mouse period1 542 mRNA expression in mouse peripheral tissues via a glucocorticoid-responsive element. J Biol 543 Chem, 280, 42036-43. 544 22. MEIJSING, S. H., PUFALL, M. A., SO, A. Y., BATES, D. L., CHEN, L. & YAMAMOTO, K. R. 2009. 545 DNA binding site sequence directs glucocorticoid receptor structure and activity. Science, 546 324, 407-10. 547 23. SCHWABISH, M. A. & STRUHL, K. (2007) The Swi/Snf complex is important for histone 548 eviction during transcriptional activation and RNA polymerase II elongation in vivo. Mol Cell 549 Biol, 27, 6987-95. 550 24. VOROBYEVA, N. E., SOSHNIKOVA, N. V., NIKOLENKO, J. V., KUZMINA, J. L., NABIROCHKINA, E. 551 N., GEORGIEVA, S. G. & SHIDLOVSKII, Y. V. (2009) Transcription coactivator SAYP combines 552 chromatin remodeler Brahma and transcription initiation factor TFIID into a single 553 supercomplex. Proc Natl Acad Sci U S A, 106, 11049-54. 554 25. WILSON, B. G. & ROBERTS, C. W. 2011. SWI/SNF nucleosome remodellers and cancer. Nat 555 Rev Cancer, 11, 481-92. 556 26. JOHN, S., SABO, P. J., JOHNSON, T. A., SUNG, M. H., BIDDIE, S. C., LIGHTMAN, S. L., VOSS, T. 557 C., DAVIS, S. R., MELTZER, P. S., STAMATOYANNOPOULOS, J. A. & HAGER, G. L. 2008. 558 Interaction of the glucocorticoid receptor with the chromatin landscape. Mol Cell, 29, 611- 559 24. 560 27. WIEGAND, K. C., SHAH, S. P., AL-AGHA, O. M., ZHAO, Y., TSE, K., ZENG, T., SENZ, J., 561 MCCONECHY, M. K., ANGLESIO, M. S., KALLOGER, S. E., YANG, W., HERAVI-MOUSSAVI, A., 562 GIULIANY, R., CHOW, C., FEE, J., ZAYED, A., PRENTICE, L., MELNYK, N., TURASHVILI, G., 563 DELANEY, A. D., MADORE, J., YIP, S., MCPHERSON, A. W., HA, G., BELL, L., FEREDAY, S., TAM, 564 A., GALLETTA, L., TONIN, P. N., PROVENCHER, D., MILLER, D., JONES, S. J., MOORE, R. A., 565 MORIN, G. B., OLOUMI, A., BOYD, N., APARICIO, S. A., SHIH, I. M., MES-MASSON, A. M., 566 BOWTELL, D. D., HIRST, M., GILKS, B., MARRA, M. A. & HUNTSMAN, D. G. 2010. ARID1A 567 mutations in endometriosis-associated ovarian carcinomas. N Engl J Med, 363, 1532-43. 568 28. ANGLESIO, M. S., WIEGAND, K. C., MELNYK, N., CHOW, C., SALAMANCA, C., PRENTICE, L. M., 569 SENZ, J., YANG, W., SPILLMAN, M. A., COCHRANE, D. R., SHUMANSKY, K., SHAH, S. P., 570 KALLOGER, S. E. & HUNTSMAN, D. G. 2013. Type-specific cell line models for type-specific 571 ovarian cancer research. PLoS One, 8, e72162 572 29. BITLER, B. G., AIRD, K. M., GARIPOV, A., LI, H., AMATANGELO, M., KOSSENKOV, A. V., 573 SCHULTZ, D. C., LIU, Q., SHIH IE, M., CONEJO-GARCIA, J. R., SPEICHER, D. W. & ZHANG, R. 574 Synthetic lethality by targeting EZH2 methyltransferase activity in ARID1A-mutated cancers. 575 Nat Med, 21, 231-8. 576 30. GUAN, B., MAO, T. L., PANUGANTI, P. K., KUHN, E., KURMAN, R. J., MAEDA, D., CHEN, E., 577 JENG, Y. M., WANG, T. L. & SHIH IE, M. 2011. Mutation and loss of expression of ARID1A in 578 uterine low-grade endometrioid carcinoma. Am J Surg Pathol, 35, 625-32. 579 31. DICKINSON, R. E., FEGAN, K. S., REN, X., HILLIER, S. G. & DUNCAN, W. C. 2011. Glucocorticoid 580 regulation of SLIT/ROBO tumour suppressor genes in the ovarian surface epithelium and 581 ovarian cancer cells. PLoS One, 6, e27792 582 32. VALLONE, D., BATTISTA, S., PIERANTONI, G. M., FEDELE, M., CASALINO, L., SANTORO, M., 583 VIGLIETTO, G., FUSCO, A. & VERDE, P. 1997. Neoplastic transformation of rat thyroid cells 584 requires the junB and fra-1 gene induction which is dependent on the HMGI-C gene product. 585 EMBO J 16:5310–5321. 586 33. SO, A. Y., BERNAL, T. U., PILLSBURY, M. L., YAMAMOTO, K. R. & FELDMAN, B. J. C. P. (2009) 587 Glucocorticoid regulation of the circadian modulates glucose homeostasis. Proc Natl 588 Acad Sci U S A, 106, 17582-7 589 34. CORNEN, S., ADELAIDE, J., BERTUCCI, F., FINETTI, P., GUILLE, A., BIRNBAUM, D. J., 590 BIRNBAUM, D. & CHAFFANET, M. 2012. Mutations and deletions of ARID1A in breast tumors. 591 Oncogene, 31, 4255-6. 592 35. GUAN, B., GAO, M., WU, C. H., WANG, T. L. & SHIH, I. M. 2012. Functional analysis of in- 593 frame indel ARID1A mutations reveals new regulatory mechanisms of its tumor suppressor 594 functions. Neoplasia, 14, 986-93. 595 36. TRAPNELL, C., WILLIAMS, B. A., PERTEA, G., MORTAZAVI, A., KWAN, G., VAN BAREN, M. J., 596 SALZBERG, S. L., WOLD, B. J. & PACHTER, L. 2010. Transcript assembly and quantification by 597 RNA-Seq reveals unannotated transcripts and isoform switching during cell differentiation. 598 Nat Biotechnol, 28, 511-5. 599 37. TRAPNELL, C., ROBERTS, A., GOFF, L., PERTEA, G., KIM, D., KELLEY, D. R., PIMENTEL, H., 600 SALZBERG, S. L., RINN, J. L. & PACHTER, L. 2012. Differential gene and transcript expression 601 analysis of RNA-seq experiments with TopHat and Cufflinks. Nat Protoc, 7, 562-78. 602 38. D’ADAMIO, F., O. ZOLLO, R. MORACA, E. AYROLDI, S. BRUSCOLI, A. BARTOLI,L. CANNARILE, 603 G. MIGLIORATI, & C. RICCARDI. 1997. A new dexamethasone- induced gene of the leucine 604 zipper family protects T lymphocytes from TCR/ CD3-activated cell death. Immunity 7, 803– 605 812. 606 39. WANG, J. C., DERYNCK, M. K., NONAKA, D. F., KHODABAKHSH, D. B., HAQQ, C. & 607 YAMAMOTO, K. R. (2004) Chromatin immunoprecipitation (ChIP) scanning identifies primary 608 glucocorticoid receptor target genes. Proc Natl Acad Sci U S A, 101, 15603-8. 609 40. WANG, S & ZHANG, S. 2011. Dickkopf-1 is frequently overexpressed in ovarian serous 610 carcinoma and involved in tumor invasion. Clin Exp , 28, 581-91. 611 41. TRAN, T., SHATNAWI, A., ZHENG, X., KELLEY, K. M. & RATNAM, M. 2005. Enhancement of 612 folate receptor alpha expression in tumor cells through the glucocorticoid receptor: a 613 promising means to improved tumor detection and targeting. Cancer Res, 65, 4431-41. 614 42. REDDY, T. E., PAULI, F., SPROUSE, R. O., NEFF, N. F., NEWBERRY, K. M., GARABEDIAN, M. J. & 615 MYERS, R. M. 2009. Genomic determination of the glucocorticoid response reveals 616 unexpected mechanisms of gene regulation. Genome Res, 19, 2163-71. 617 43. GEORGE, C. L., BIRNIE, M.T., FLYNN, B.P., KERSHAW, Y.M., LIGHTMAN, S. L. & CONWAY- 618 CAMPBELL, B. L. 2016. Ultradian glucocorticoid exposure directs gene-dependent and tissue- 619 specific mRNA expression patterns in vivo. Mol Cell Endcrinol, 439, 46-53. 620 44. REDDY, T. E., GERTZ, J., CRAWFORD, G. E., GARABEDIAN, M. J. & MYERS, R. M. 2012. The 621 hypersensitive glucocorticoid response specifically regulates period 1 and expression of 622 circadian genes. Mol Cell Biol, 32, 3756-67. 623 45. DELBOVE, J., ROSSON, G., STROBECK, M., CHEN, J., ARCHER, T. K., WANG, W., KNUDSEN, E. S. 624 & WEISSMAN, B. E. 2011. Identification of a core member of the SWI/SNF complex, 625 BAF155/SMARCC1, as a human tumor suppressor gene. Epigenetics, 6, 1444-53. 626 46. GUAN, B., WANG, T. L. & SHIH, I. M. 2011. ARID1A, a factor that promotes formation of 627 SWI/SNF-mediated chromatin remodeling, is a tumor suppressor in gynecologic cancers. 628 Cancer Res, 71, 6718-27. 629 47. CHANDLER, R. L., DAMRAUER, J. S., RAAB, J. R., SCHISLER, J. C., WILKERSON, M. D., DIDION, J. 630 P., STARMER, J., SERBER, D., YEE, D., XIONG, J., DARR, D. B., PARDO-MANUEL DE VILLENA, F., 631 KIM, W. Y. & MAGNUSON, T. 2015. Coexistent ARID1A-PIK3CA mutations promote ovarian 632 clear-cell tumorigenesis through pro-tumorigenic inflammatory cytokine signalling. Nat 633 Commun, 6, 6118. 634 48. YAMAMOTO, S., TSUDA, H., TAKANO, M., TAMAI, S. & MATSUBARA, O. 2012. PIK3CA 635 mutations and loss of ARID1A protein expression are early events in the development of 636 cystic ovarian clear cell adenocarcinoma. Virchows Arch, 460, 77-87. 637 49. SAMARTZIS, E. P., NOSKE, A., DEDES, K. J., FINK, D. & IMESCH, P. 2013. ARID1A mutations 638 and PI3K/AKT pathway alterations in endometriosis and endometriosis-associated ovarian 639 carcinomas. Int J Mol Sci, 14, 18824-49. 640 50. ARANCIBIA, S., BENÍTEZ, D., NÚÑEZ, L. E., JEWELL, C. M., LANGJAHR, P., CANDIA, E., ZAPATA- 641 TORRES, G., CIDLOWSKI, J. A., GONZÁLEZ, M. J. & HERMOSO, M. A. 2011. 642 Phosphatidylinositol 3-kinase interacts with the glucocorticoid receptor upon TLR2 643 activation. J Cell Mol Med, 15, 339-49. 644 51. MURATCIOGLU, S., PRESMAN, D. M., POOLEY, J. R., GRØNTVED, L., HAGER, G. L., NUSSINOV, 645 R., KESKIN, O. & GURSOY, A. 2015. Structural Modeling of GR Interactions with the SWI/SNF 646 Chromatin Remodeling Complex and C/EBP. Biophys J, 109, 1227-39 647 52. CHEN, J. & ARCHER, T. K. 2005. Regulating SWI/SNF subunit levels via protein-protein 648 interactions and proteasomal degradation: BAF155 and BAF170 limit expression of BAF57. 649 Mol Cell Biol, 25, 9016-27. 650 53. LIANG, X., GUO, Y., FIGG, W. D., FOJO, A. T., MUELLER, M. D. & YU, J. J. 2011. The Role of 651 Wild-Type p53 in Cisplatin-Induced Chk2 Phosphorylation and the Inhibition of Platinum 652 Resistance with a Chk2 Inhibitor. Chemother Res Pract, 2011, 715469. 653 54. SON, D. S., KABIR, S. M., DONG, Y. L., LEE, E. & ADUNYAH, S. E. 2012. Inhibitory effect of 654 tumor suppressor p53 on proinflammatory chemokine expression in ovarian cancer cells by 655 reducing proteasomal degradation of IκB. PLoS One, 7, e51116. 656 55. LIANG, H., KOWALCZYK, P., JUNCO, J. J., KLUG-DE SANTIAGO, H. L., MALIK, G., WEI, S. J. & 657 SLAGA, T. J. 2014. Differential effects on lung cancer cell proliferation by agonists of 658 glucocorticoid and PPARα receptors. Mol Carcinog, 53, 753-63 659 56. CHAN-ON, W., NAIRISMÄGI, M. L., ONG, C. K., LIM, W. K., DIMA, S., PAIROJKUL, C., LIM, K. H., 660 MCPHERSON, J. R., CUTCUTACHE, I., HENG, H. L., OOI, L., CHUNG, A., CHOW, P., CHEOW, P. 661 C., LEE, S. Y., CHOO, S. P., TAN, I. B., DUDA, D., NASTASE, A., MYINT, S. S., WONG, B. H., GAN, 662 A., RAJASEGARAN, V., NG, C. C., NAGARAJAN, S., JUSAKUL, A., ZHANG, S., VOHRA, P., YU, W., 663 HUANG, D., SITHITHAWORN, P., YONGVANIT, P., WONGKHAM, S., KHUNTIKEO, N., 664 BHUDHISAWASDI, V., POPESCU, I., ROZEN, S. G., TAN, P. & TEH, B. T. 2013. Exome 665 sequencing identifies distinct mutational patterns in liver fluke-related and non-infection- 666 related bile duct cancers. Nat Genet, 45, 1474-8 667 57. CHANG, C., NORRIS, J. D., GRØN, H., PAIGE, L. A., HAMILTON, P. T., KENAN, D. J., FOWLKES, 668 D. & MCDONNELL, D. P. 1999. Dissection of the LXXLL -coactivator 669 interaction motif using combinatorial peptide libraries: discovery of peptide antagonists of 670 estrogen receptors alpha and beta. Mol Cell Biol, 19, 8226-39. 671

672

673 Highlights (for review)

Highlights

SKOV3 cells have a truncated ARID1a SKOV3 cells exhibit a highly restrictive response to glucocorticoids We have identified impaired GR binding at a chromatin remodelling inducible site in the highly inducible glucocorticoid target gene, Per1. Consistent with reduced GR binding, Per1 shows an impaired and delayed transcriptional induction. Supplementary Material Click here to download Supplementary Material: STUBBS et al 2017 supplementary.docx